PDF Publication Title:
Text from PDF Page: 137
6SOFTWARE If you build it, he will come. Field of Dreams Over the course of this thesis, we developed a few software packages to aid our work. All are publicly available. This chapter briefly describes each package and its major architecture and design goals. A complete description of each package would be nearly as lengthy as the current document and needlessly tedious. Instead, we have editorialized and highlighted the most interesting or challenging pieces of each implementation. One of the recurring themes in all the packages is implementing graph structures in Matlab. We discuss the setting for these idea in sections 6.1 and 6.2. The MatlabBGL package (section 6.3) is an interface between the Boost graph library [Siek et al., 2001] and Matlab.1 The interface is based on the native Matlab sparse matrix and provides a rich suite of graph algorithms that were previously unavailable in Matlab. Currently, the suite works on Windows, Mac OSX, and Linux in both 32-bit and 64-bit computing envi- ronments. It has been downloaded over 7,000 times and used in numerous publications. Whereas MatlabBGL is a complicated set of C and C++ routines to extract the maximum performance for graph operations in Matlab, the gaimc rou- tines (section 6.4) are written in pure Matlab m-code.2 These algorithms are slower than their MatlabBGL counterparts, although, they are faster than the dictum “for-loops in Matlab are really slow” suggests. For a few algorithms, the gaimc code runs in twice the MatlabBGL time; for others, it runs in slightly more than four times the MatlabBGL time. The benefit of the library is that it is easy to add new algorithms with these performance characteristics. Adding algorithms to MatlabBGL is quite difficult. In this thesis, many of the graphs explored are extremely large. When these are graphs created by crawling the WWW, they have some regular- ity that makes them highly compressible. Boldi and Vigna [2005] develop coding schemes to compress these graphs at under three bits per edge. Re- cent improvements in applying the same technique enhance compression to around one bit per edge [Boldi et al., 2009]. Our libbvg code (section 6.5) re-implements pieces of the bvgraph framework in portable C code to enable a Matlab wrapper and a shared memory parallel interface. Writing a true Python wrapper for the graphs should also be possible. Many of the contributions of this thesis have already been published. Each publication has an accompanying software package providing all the source code and experiment scripts. These packages are briefly described in sec- tion 6.6. 117 1 In the remainder of this chapter, we are going to remove the special styling on Matlab. Matlab is used as a proper noun throughout the remainder of the chapter instead of a product name. 2 Matlab m-code is how we refer to Matlab’s native language.PDF Image | MODELS AND ALGORITHMS FOR PAGERANK SENSITIVITY
PDF Search Title:
MODELS AND ALGORITHMS FOR PAGERANK SENSITIVITYOriginal File Name Searched:
gleich-pagerank-thesis.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com (Standard Web Page)